AI Agent Hub
Back to plugins
dsh-team-superflow preview

dsh-team-superflow

Client Updated 2026.08.26

Run the following command in DeepSeek Harness:

dsh plugin install fengbin01/dsh-team-superflow

Paste the following prompt into your AI chat to install this plugin:

Run dsh plugin install fengbin01/dsh-team-superflow in DeepSeek Harness to install it, with the full source available at https://github.com/fengbin01/dsh-team-superflow

About this plugin

Every time you spin up multiple Agents in DeepSeek Harness AgentTeams, you are likely to hit the same pain points: vague role boundaries, ad-hoc workflows, and no real visibility into progress. dsh-team-superflow fixes this by registering a stable five-member dev team—spec writer, architect, engineer, reviewer, and docs/releaser—and driving the entire engagement through the eight-state spec-superflow state machine, so each step has a clear phase, entry conditions, and exit criteria.

The state machine does the heavy lifting. When every task in the current phase completes, the team auto-advances to the next one. Once the work reaches the approved-for-build state, the model is locked out and a human must explicitly click the approval button in the Web panel before implementation can proceed, creating a hard build gate. If scope shifts or contract drift is detected, the captain can rewind the team back to specifying or bridging. Every task is tagged with a phase, and the scheduler only dispatches tasks that match both the current phase and the member role scope. In the activity panel, a stepper of eight phase pills, an overall percentage, and per-phase task counts provides a live, glanceable progress view grounded in the durable host state.

This plugin is a good fit for DeepSeek Harness users who want structured, repeatable multi-Agent collaboration with human oversight at critical checkpoints and visual progress tracking. If loose, role-ambiguous, gate-free teamwork has become a drag on your workflow, the fixed roster and state-machine discipline offered here should feel like a welcome upgrade.

Screenshots

Use Cases

  • Automatically register a fixed five-member dev team and drive it through the eight-state workflow on every AgentTeams run
  • Freeze implementation at the build gate and approve it manually in the Web panel before merging
  • Rewind the team to specifying or bridging when scope changes or contract drift is detected

Best For

  • Dev teams that rely on multi-Agent collaboration in DeepSeek Harness on a regular basis
  • Quality-sensitive users who want human approval at critical checkpoints rather than model self-confirmation
  • Users who prefer structured workflows and real-time progress tracking over ad-hoc role assignment